Keynes's analysis of agents' behaviour under uncertainty is highly original and fundamentally different from the conventional approach that fails to distinguish between risk and uncertainty. The analytical representation of the informational environment facing decision makers is a first point of difference. In contrast to the standard approach, which follows the method of classical physics in assuming that outcomes of human actions are independent of their dates, Keynesian theory leads to the conclusion that economics has to deal with actions in historical time when random functions are not in a state of statistical control. This conclusion exposes a divergence of objectives between the two approaches. The axiomatic construction of neoclassical theory seeks to derive optimal states and behaviours. The Keynesian approach takes the opposite tack and seeks to determine the informational environment for decision making on the basis of observations taken from the evolution of the 'real world'. The identification of the two characteristic features of this approach in historical time and crucial decisions makes it possible for Keynesian economists to identify the chronic inability of standard theory to deal with conditions of fundamental uncertainty that may be defined in a way that is radically different from the familiar notion of risk. It also makes it possible to take into account a more complex economic environment based on non-ergodic processes in which decisions once taken will evolve in an indeterminate way.
In a non-stationary (non-ergodic) environment it is possible to highlight a second point of difference from neoclassical approaches, and more particularly from the New Classical Economists.11 This point is associated with the elaboration by Keynesian economists of a conception of 'rationality' that is fundamentally different from that characterized by the 'robot mentality' of the theory of rational expectations. This type of rationality, which is capable of creating 'expectational instability', has little to do with 'irrational psychological fluctuations'. It is not based on the idea that, once uncertainty is taken into account, it is no longer possible to make theoretical statements about initial conditions and results. Rather, it leads to the identification of certain specific rational responses to uncertainty that may be identified and formalized. If, in fact, the notion of uncertainty calls into question the traditional approach to rationality, it can nonetheless be expressed in the form of 'law-like' rules of behaviour. It is thus possible to submit it to scientific analysis and debate. This conclusion applies in particular to the notion of 'rational spirits' that is the foundation of the Keynesian analysis of behaviour under fundamental uncertainty. 'Rational spirits' cannot be interpreted as the result of the cognitive irrationality of agents, as they are rather based on the notion of logical probability developed by Keynes in his TP, as well as on the 'conventional individual rationality' discussed in GT and further investigated in the 1937 Quarterly Journal of Economics article.